Nejumi 4 - jaster (2-shot) - MMLU-ProX (Japanese): leaderboard
the Japanese split of MMLU-ProX, a multilingual version of MMLU-Pro, from the llm-jp-eval jaster suite as run 2-shot by Nejumi Leaderboard 4 (Weights & Biases Japan). Score is exact match (%) on Nejumi's sampled test split; higher is better.
Metric: Exact match (%). Source: nejumi.ai. Saturation forecast: Estimated already saturated. 145 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Hy4 preview (High) | 92 | #45 (Hy4 preview) |
| 2 | Gemini 3.1 Pro (Preview) | 91 | #64 |
| 3 | Gemini 3 Flash (Preview) | 91 | #90 |
| 4 | Gemini 3.6 Flash | 91 | #57 |
| 5 | GLM-5.3 (Max) | 91 | #44 (GLM-5.3) |
| 6 | Qwen 3.6 Max (Preview) (Thinking) | 91 | |
| 7 | Kimi K3 (Thinking) | 91 | #35 (Kimi K3) |
| 8 | Gemini 3 Pro (Preview) | 90 | #71 |
| 9 | Gemini 3.5 Flash | 90 | #66 |
| 10 | GPT-5.6 Sol (Max) | 90 | #22 (GPT-5.6 Sol) |
| 11 | Kimi K2.5 (Thinking) | 90 | #152 (Kimi K2.5) |
| 12 | Claude Opus 5 (Max) | 90 | #16 (Claude Opus 5) |
| 13 | Grok 4.5 (High) | 90 | #54 (Grok 4.5) |
| 14 | Muse Spark Open (xHigh) | 90 | #42 (Muse Spark Open) |
| 15 | Qwen 3.5 397B A17B (Thinking) | 90 | #154 (Qwen 3.5 397B A17B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=nejumi-4-jaster-2-shot-mmlu-prox-japanese · How It Works · Data refreshed daily, snapshot 2026-11-05.